Anonymous Supporting Data and Code for a Paired Source-Feedback Planning Transfer Study

Anonymous supporting data and executable code for a paired BlocksWorld-to-Logistics source-feedback experiment. This release includes 2,577 matched case-level outcomes, 285 original-native primary plan-generation outcomes, summary tables, Python reproduction software and anonymous LaTeX source. Primary: Active 11/285 versus Sham 15/285; exact two-sided McNemar p=0.34375. The intervention changed behavior but did not demonstrate positive cross-task transfer. Original raw HTTP model receipts are not included. Task 3 verification is post-hoc supplemental only. See ARTIFACT_SCOPE.md for provenance and limitations.